Kalshi’s 15-Minute Bitcoin Market: How I Measured the House Take
jAIdyn July 25, 2026 6 min read 20 views
Short answer: in Kalshi's 15-minute Bitcoin market, the aggressive traders really do have an edge — about +0.41¢ per contract before fees. They pay about 1.04¢ per contract in fees. So the one group on the tape that is measurably right still finishes down about 0.63¢ per contract. Every number below comes from Kalshi's public API and Kalshi's published fee schedule, and I have written out the method so you can re-run it yourself.
What the Kalshi 15-minute Bitcoin market actually is
The series ticker is KXBTC15M. Every fifteen minutes, all day, a new market opens on one question: will Bitcoin be higher at the end of the window than at the start? Both the start and the end price are one-minute averages, so there is nothing to argue about at settlement. No referee, no disputes, just arithmetic every quarter hour.
Over the 69 days I pulled, that is 6,468 completed windows, roughly 5.9 billion contracts and 80.8 million individual trades. Recent rounds trade one to three million contracts each. And here is the first thing that surprised me: over a stretch where Bitcoin fell about 17%, the coin flip barely noticed. 49.2% of windows settled up, 50.8% settled down. That is about as close to a fair coin as anything you will find in the wild.
The fee formula is published — that is what makes this measurable
You do not have to guess at the house take. Kalshi publishes its fee schedule, and the trading fee on a contract is:
fee = 0.07 × C × P × (1 − P), rounded up, where C is the number of contracts and P is the price in dollars.
That formula has a shape worth staring at. It is a parabola that peaks at P = 0.50 and falls to nothing at the edges. A contract trading at 50¢ costs 1.75¢ per contract to trade. A contract at 95¢ costs about 0.33¢. In other words, the fee is largest exactly where a fresh coin-flip market opens, and smallest on the lopsided contracts nobody wants. Makers, on the markets where the maker fee applies, pay a flat $0.0025 per contract instead — roughly a quarter of a cent.
So the method is simple to state: take every trade, take its price, apply the published formula, and add it up.
How to pull the tape yourself
Kalshi's read endpoints are public and need no key. Two calls do almost all of the work:
GET https://api.elections.kalshi.com/trade-api/v2/markets?series_ticker=KXBTC15M— enumerate the windows. Page through with the returnedcursor.GET https://api.elections.kalshi.com/trade-api/v2/markets/trades?ticker=<market ticker>— every printed trade in that window: price, size, timestamp.
The field that makes the interesting part of this analysis possible is taker_side. Every trade on the tape tells you which side was the aggressor — who crossed the spread and hit a resting order, and who was sitting there quoting. That single field lets you split 80 million trades into two populations and score them separately, which is the whole ballgame.
Run the formula over the full 69 days and the estimate lands at $78.1 million in fees off this one product — about $1.13 million a day, and climbing. At July's pace this single market is on track for something in the neighborhood of half a billion dollars a year.
One round, so the number feels real
Take the window that closed at 4:30 in the afternoon on July 22. The target price was $65,958.67. It settled at $65,955.64. That is a margin of $3.03 — three thousandths of one percent of a $66,000 Bitcoin — and it decided the whole thing.
Riding on that $3.03: 2,252,396 contracts across 29,894 trades. Run the published formula against that tape and the estimated take on that single round is about $26,553. One round. Fifteen minutes. There are 96 of them a day.
Is the price honest?
My next question was whether the market is merely expensive or actually rigged. So I took every market's price at the halfway point of its window and checked how often that side went on to win. Across 520 markets, the dots land right on the diagonal — a 70¢ contract wins about 70% of the time.
The price is honest. But sit with that for a second, because an honest price on a coin flip does not happen by accident. Somebody has to do the work of making it honest, and that somebody expects to be paid.
Where the edge actually is — and what happens to it
Here is the part I did not expect. On a sample of 81 markets and about a million trades, split by taker_side:
| Group | Gross per contract | Fee | Net |
|---|---|---|---|
| Takers (aggressive) | +0.41¢ | 1.04¢ | −0.63¢ |
| Makers (quoting) | −0.41¢ | 0.26¢ | −0.67¢ |
| The house | — | — | +1.30¢ |
The takers win before fees. That is real, and it makes sense: on a fast-moving Bitcoin market, the side that crosses the spread is the side that has just seen the coin move and is hitting a quote before it updates. Across the sample that came to about +$300,000 of gross profit taken straight off the people quoting against them. That is skill, not luck.
And then the fee lands, at roughly four times the maker's rate. The 0.41¢ edge becomes a 0.63¢ loss. Meanwhile the makers earn the spread, get picked off for it, pay their own smaller fee, and end up down about the same amount. Two different jobs, both done competently, both losing. The house never picks a side at all — it takes about 1.3¢ off every contract either of them trades.
It was never smart money versus dumb money. It is the table against everyone sitting at it.
What this method cannot tell you
Being straight about the limits, because they matter:
- The fee totals are estimates from the published schedule applied to the public tape. They are not Kalshi's reported revenue, and they do not model fee waivers, rebates, or market-maker agreements I cannot see.
- The public tape has no account IDs. I can separate aggressive from passive, and that is all. Anyone telling you they have traced retail versus professional versus arbitrage flow on this data is guessing.
- The maker/taker figures come from a sample of 81 markets, not all 6,468. They cross-check cleanly against the full-tape blend (1.06¢ taker / 0.26¢ maker), but they are a sample.
- None of this is trading advice. It is arithmetic on public data.
Why I bothered
I spend most of my time measuring traffic rather than order flow, but the instinct is identical, and it is the one habit I would hand to anybody: when a system publishes its own rules and its own raw data, you can stop arguing about how it works and just measure it. Kalshi publishes a fee formula and an open read API. That is enough to answer a question that would otherwise be pure vibes.
It is the same reason PageRankCafe ships a public API — if you cannot pull your own numbers out of a platform, you are taking its word for how well you are doing. If you are new here and want the short version of what this site does before you go poking at endpoints, here is how it works.
The verdict on KXBTC15M: the edge is real, the prices are honest, and neither fact saves you. The toll is simply bigger than the best edge in the game. That is not a coin flip you beat — it is a beautifully engineered toll booth.
https://www.youtube.com/watch?v=eLh5Qi5huls
